Safely escalating the radiation dose to locally-advanced non-small cell lung cancer patients to improve local tumor control and survival remains a challenge due to the numerous types and large magnitude of day-to-day geometric change in these patients. Deformable image registration (DIR) has emerged as a powerful tool for day-to-day mapping of target and risk structures and assessing the cumulative delivered dose to guide adaptation of the treatment plan to mitigate geometric changes. However, change in tissue and organ character - tissue formation, disintegration, and change of tissue condition between healthy and diseased - during the treatment course violates the fundamental principles upon which current DIR algorithms are based. Tissue character changes are common in lung cancer, due to primary tumor and involved lymph node mass changes in response to therapy and formation, resolution, or progression of partially collapsed lung, pleural fluid, and other associated pathologies. Tissue character changes are the type of geometric change most susceptible to delivery errors, the most likely to benefit from plan adaptation, and the most common site of failure of current DIR algorithms. Consequently, failure of DIR introduces error into the adaptive radiotherapy process, limiting the ability to safely escalate the radiation dose. We propose to develop and evaluate a new type of DIR algorithm designed to perform accurately under the difficult conditions of tumor and lung tissue character change. The algorithm is specifically designed to identify and register regions of anatomy consistent in shape and tissue character from day to day, while ignoring regions of variable character (such as collapsed lung and tumor) in the registration, and reconstructing deformation in the pathological regions using the measured deformation of the adjacent, consistent anatomical features. The project will be accomplished in three specific aims. In specific aim 1, consistent anatomy registration will be developed and validated against a database of fan beam CT images of breath hold and free-breathing lung cancer patients undergoing image- guided radiation therapy. In specific aim 2, we will extend consistent anatomy registration to cone beam CT images to enable rapid estimation of tissue location and delivered dose with the patient in the treatment position. In specific aim 3, the impact of DIR error on the ability to accurately adapt the treatment plan will be measured for consistent anatomy registration and for conventional DIR, for a variety of forms of adaptive radiotherapy. This project will provide the radiation therapy community with accurate DIR to support adapting the radiation therapy treatment plan to geometric change in locally-advanced non-small cell lung cancer.